You still need a great deal of processing for the signal and image stabilization/readings from the sensors. This test version gets away with the M0 for that (well, as much as one can say that looking at the lens) because it only has ~61k of on/off pixels to make a HUD, which also doesn't need to be refreshed often just restabilized locally. Typical AR headsets get away with not having to deal with stabilizing eye movements, more place to stuff processing anyways, and no concern on being able to receive massive bandwidth video streams for more lively content.